No, because you should assume all code will be analyzed and attacked by adversarial AIs. That’s the real world impact of Project Glasswing and the like. If attackers are using it so should the blue team. AI analysis should be a part of your security review but not the whole thing.

Updated parent comment. Ideally, looking beyond this work, and more generally, a funded AI would be used to do analysis and then to dispatch tasks to qualified humans. A network of available qualified humans would have to exist that the AI can access. Humans could then of course provide feedback to AI for the loop to continue with new tasks to humans. Think Uber but more generally for AI to tap into real-world work and expertise.